From the course: Creating Agents with Python and the AI Toolkit for Visual Studio Code

Unlock this course with a free trial

Join today to access over 25,200 courses taught by industry experts.

Using trace view to inspect step-by-step agent behavior

Using trace view to inspect step-by-step agent behavior

From the course: Creating Agents with Python and the AI Toolkit for Visual Studio Code

Start my 1-month free trial Buy for my team

Using trace view to inspect step-by-step agent behavior

“

So let's now run this agent and start collecting traces. So far, I have not saved GitHub Copilot's changes yet. We'll wait to see how it does. To begin, what I will do is run the command Python sales-insite-agent.py. What it should do from here is run the agent file as it has before. If there are any errors, fortunately, we could leverage GitHub Copilot to help us with resolving those. But from what seems like, things are working as we'd hope. We do have some tool calls here. I'm seeing the prompts are being passed in too. Once we are done here, we see all tasks were completed successfully. We can come to this tracing viewer that we saw earlier, and I do see that we have some traces, which is great. The most recent two traces were from the run that we just did. If I scroll over to the far right, I can see the duration of the traces. The very first one took 12 seconds and the next one took five seconds. I can also see the total tokens that were used in totality. But what's in a trace?…

Contents